When: 3rd November 2021
Checklist:
- Create a task to communicate the chosen date and send an announcement to the community: T293965
- Create a calendar entry for the maintenance, invite sre-data-persistence@
- Add to deployments calendar. E.g.:
{{Deployment calendar event card |when=2021-11-2 22:00 SF |length=0.5 |window=Database primary switchover for s1 |who={{ircnick|kormat|Stevie Beth Mhaol}}, {{ircnick|marostegui|Manuel 'Early Bird' Arostegui}} |what=https://phabricator.wikimedia.org/T293964 }}
NEW primary: db1118
OLD primary: db1163
- Check configuration differences between new and old primary:
sudo pt-config-diff --defaults-file /root/.my.cnf h=db1163.eqiad.wmnet h=db1118.eqiad.wmnet
Failover prep:
- Silence alerts on all hosts:
sudo cookbook sre.hosts.downtime --hours 1 -r "Primary switchover s1 T293964" 'A:db-section-s1'
- Set NEW primary with weight 0
sudo dbctl instance db1118 set-weight 0 sudo dbctl config commit -m "Set db1118 with weight 0 T293964"
- Topology changes, move all replicas under NEW primary
sudo db-switchover --timeout=15 --only-slave-move db1163 db1118
- Disable puppet on both nodes
sudo cumin 'db1163* or db1118*' 'disable-puppet "primary switchover T293964"'
- Merge gerrit puppet change to promote NEW primary: https://gerrit.wikimedia.org/r/736114
Failover:
- Log the failover:
!log Starting s1 eqiad failover from db1163 to db1118 - T293964
- Set section read-only:
sudo dbctl --scope eqiad section s1 ro "Maintenance until 06:15 UTC - T293964" sudo dbctl config commit -m "Set s1 eqiad as read-only for maintenance - T293964"
- Check s1 is indeed read-only
- Switch primaries:
sudo db-switchover --skip-slave-move db1163 db1118 echo "===== db1163 (OLD)"; sudo mysql.py -h db1163 -e 'show slave status\G' echo "===== db1118 (NEW)"; sudo mysql.py -h db1118 -e 'show slave status\G'
- Promote NEW primary in dbctl, and remove read-only
sudo dbctl --scope eqiad section s1 set-master db1118 sudo dbctl --scope eqiad section s1 rw sudo dbctl config commit -m "Promote db1118 to s1 primary and set section read-write T293964"
- Restart puppet on both hosts:
sudo cumin 'db1163* or db1118*' 'run-puppet-agent -e "primary switchover T293964"'
Clean up tasks:
- Clean up heartbeat table(s).
- change events for query killer:
events_coredb_master.sql on the new primary db1118 events_coredb_slave.sql on the new slave db1163
- Update DNS: https://gerrit.wikimedia.org/r/c/operations/dns/+/736115/
- Update candidate primary dbctl notes
sudo dbctl instance db1163 set-candidate-master --section s1 true sudo dbctl instance db1118 set-candidate-master --section s1 false
- Check tendril was updated
- Check zarcillo was updated
- db-switchover should do this. If it fails, do it manually: https://phabricator.wikimedia.org/P13956
- Depool OLD primary, as it's running 10.1, replicating from a 10.4 primary
sudo dbctl instance db1163 depool sudo dbctl config commit -m "Depool db1163 until it's reimaged to buster T293964"
- Apply outstanding schema changes to db1163 (if any) NONE
- Update/resolve this ticket.